Precinct Level Exploration of Civilian Complaints Against New York City Police Officers

August 12, 2020 - Joy Yang

This July, Propublica released a database of over 12,000 closed allegations filed against NYPD officers with at least one substantiated allegation, and are still on the force as of June 2020.

While an officer-level analysis would be most interesting due to the nature of this data, because of the accusatory tone that such an analysis could easily take, we’ll, for now, focus on a precinct-level analysis of the data, and possibly return to an officer-level analysis at a later time.

One immediate question that comes to mind in considering precinct-level analyses is: “Do any precincts have more allegations than would be expected? Do any precincts have less allegations than would be expected?”

To answer this question, we must first define what is “expected.” As discussed in J124, if not normalized thoughtfully, choropleth maps can become simply representations of population density. To account for this, one approach might be to calculate the number of allegations per month per 100,000 residents; however, this metric highlights one particular precinct: Central Park. While in the last 10 years, there have only been 5 allegations issued from Central Park, because only 25 people claim to live within the precinct, the park appears to have an abnormally high number of rate of police allegations.

It appears, then, that the number of residents may then not be the most appropriate normalization factor. Instead, a more suitable factor could be the number of complaints made to the NYPD about crimes/suspicious activities within each precinct. This captures, in essence, a crime rate. Normalizing allegations about the NYPD by the complaints made to the NYPD would allow us to compare allegation rates given the frequency that police are called upon in various precincts.

And in order to keep the analysis more recent, only allegations made in the last 10 years (and complaints made to the NYPD in the last 10 years) are included in the visualization below.

It should be noted that because of the nature of this data, there is a major caveat: complaints made within the last 10 years against officers no longer in the NYPD will not be included in this dataset.

This map implies that there seems to be a spatial structure behind police allegation rates. Is this due to the nature of the New York neighborhoods/zoning? Or is this due to the police culture, or even individual police who may have high/low allegation rates? These questions will need further exploration through an officer-level analysis.

However, another question that arises is, “What types of allegations are common? Do allegation types follow a geographic structure as well?”

Overall, for this dataset, in the last 10 years, the most common allegations are:

top_allegations
Force: Physical force 16%
Discourtesy: Word 13%
Abuse of Authority: Premises entered and/or searched 8%
Abuse of Authority: Stop 7%
Abuse of Authority: Search (of person) 5%

We can additionally examine the most common allegation for each precinct, and scale the opacity according to the prevalence of the most common allegation as follows:

From the map above, there is some suggestion that allegation types may follow an underlying geographic structure. This again, this brings up the question of whether this may be due to the characteristics of neighborhood and crimes reported for that precinct, or whether this may be due to police behavior and the policing culture of the precinct.

This leaves many interesting further directions for exploration, a few additional questions we can answer with this data include, for example: